Silent Speech Interface Using Lip-Reading Methods

A silent speech interface (SSI) maps articulatory movement data to forecast the speech of mute people. SSIs have vast potential for improving oral communication in people who have had laryngectomy surgery or those with severe voice abnormalities. In this paper, we present a lip-reading recognition algorithm to recognize English words from the lips when speaking. The lip contour and English words were distinguished using a variety of lip metric metrics, including width, height, contour points, area, and ratio (width/height). In this test, fifteen talkers mouthed (silently articulated) sentences to finish a phrase-reading assignment in an experimental setting. Among the fifteen participants, 89% and 81% of the mouthed phrases were correctly recognized for two different classifier cases respectively. The automated ROI detection, elimination of lip shape effects, and operation without prior data training are strengths of the proposed method. The experimental outcomes showed the viability and potential of lip-reading-based silent speech interfaces for future clinical applications.